Corrected score methods for estimating Bayesian networks with error-prone nodes.

Area of Science:

• Computational biology
• Network inference
• Statistical modeling

Background:

• Cellular signaling networks are crucial for understanding biological processes.
• Flow cytometry generates high-dimensional data but is prone to measurement error.
• Accurate network inference is challenging with noisy biological data.

Purpose of the Study:

• To develop robust methods for inferring Bayesian networks from error-prone data.
• To address challenges in cellular signaling network inference using flow cytometry.
• To improve the accuracy of causal relationship identification in biological networks.

Main Methods:

• Proposed penalized estimation methods to infer causal relationships in Bayesian networks.
• Incorporated measurement error correction and sparsity encouragement into network estimation.
• Developed a tuning parameter selection approach for penalized estimation.
• Compared proposed methods against a naive approach ignoring measurement error.

Main Results:

• Demonstrated improved accuracy in network inference by accounting for measurement error.
• Showcased the effectiveness of penalized estimation for sparse network recovery.
• Empirical studies validated the proposed methods against a naive approach.
• Successfully applied the methods to infer single-cell flow cytometry data.

Conclusions:

• The developed methods provide a more accurate approach to inferring cellular signaling networks.
• Accounting for measurement error is critical for reliable network inference from flow cytometry data.
• These techniques enhance the understanding of biological signaling pathways through robust network analysis.